lak15 twitter archeology

31
TWITTER ARCHEOLOGY OF LEARNING ANALYTICS AND KNOWLEDGE CONFERENCES Bodong Chen, University of Minnesota Xin (Cindy) Chen, Purdue University Wanli Xing, University of Missouri #LAK15, Marist College, Poughkeepsie, NY, March 20, 2015 Authors first met at the LAK14 Doctoral Consortium. Thanks, LAK! @bodong_c @magic_cindy @helloworld_xing #LAK15Meta

Upload: bodong-chen

Post on 16-Jul-2015

395 views

Category:

Data & Analytics


0 download

TRANSCRIPT

TWITTER ARCHEOLOGY OF LEARNING ANALYTICS AND KNOWLEDGE CONFERENCES

Bodong Chen, University of MinnesotaXin (Cindy) Chen, Purdue UniversityWanli Xing, University of Missouri

#LAK15, Marist College, Poughkeepsie, NY, March 20, 2015Authors first met at the LAK14 Doctoral Consortium. Thanks, LAK!

@bodong_c@magic_cindy@helloworld_xing

#LAK15Meta

Source: http://eeecos.org/

Why Twitter at Conferences?

#LAK15Meta

1. When did you first tweet about #LAK?2. Why/What do you tweet (e.g., comments, info, beer)?3. Who did you get to know through #LAK?4. What is your primary research topic?

Why “Twitter Archeology”?

● Not all have published yet● Not all are interested in publishing● Broader participation & richer

interactions (cf. citing)

#LAK15Meta

#LAK15Meta

#LAK15Meta

Questions

● Did Twitter enable participation and conversation?

● Was participation persistent?● Social dynamics & change over time?● Underlying topics & change over time?

#LAK15Meta

(Cleaned) Dataset

Conference Participants Tweets

LAK11 215 1358

LAK12 606 4050

LAK13 280 2223

LAK14 362 3105

* Data wrangling challenges: inconsistencies of data shapes across years; a systematic mistake of user ids in the 2011 archive; parsing interactions; etc.

3587 (by last night) LAK15 465

(Cleaned) Dataset

Conference Participants Tweets

LAK11 215 1358

LAK12 * 606 4050

LAK13 280 2223

LAK14 362 3105

* LAK12: “A substantial amount of tweets during LAK12 was about the technologies adopted for live video streaming.”

#LAK15Meta

An Overview of Analyses

● Descriptive Analysis● “Flow” of Twitter Participants● Interaction Social Networks● Evolution of Topics

#LAK15Meta

Descriptive: Conferences

Conference Tweets Retweets Replies

LAK11 1358 450 (33.1%) 230 (16.9%)

LAK12 4050 1207 (29.8%) 430 (10.6%)

LAK13 2223 570 (25.6%) 363 (16.3%)

LAK14 3105 1255 (40.4%) 570 (18.4%)

#LAK15Meta

Descriptive: Individuals

Outgoing Incoming

Conf Tweets Retweets Replies Retweets Replies

LAK11 6.3 (14.1) 2.1 (4.7) 1.1 (3.5) 2.0 (7.5) 0.9 (3.4)

LAK12 6.7 (23.8) 2.0 (5.0) 0.7 (2.9) 1.9 (13.4) 0.6 (3.7)

LAK13 7.9 (31.5) 2.0 (4.5) 1.3 (7.4) 2.0 (7.7) 1.1 (4.2)

LAK14 8.6 (30.9) 3.5 (10.4) 1.6 (7.5) 3.5 (13.5) 1.5 (6.0)

Means and standard deviations of activities of individuals

#LAK15Meta

“Flow” of Participants#LAK15Meta

* The only reason you see a pie here is we just celebrated a big Pie Day – 3.14.15 ;)

1,217 unique participants

Peripheral participation

#LAK15Meta

# of years of participation

Interaction Networks based on retweets,

replies and mentions

* The node size and color are based on betweenness centrality.

Interaction Networks

Conf Nodes Edges Avg Degree

Avg Path Length

Reciprocate Rate

# of Communities

LAK11 215 569 2.65 2.95 .13 3

LAK12 606 1521 2.51 3.1 .13 6

LAK13 280 736 2.63 2.99 .15 5

LAK14 362 1369 3.78 2.73 .20 6

#LAK15Meta

Content: Hashtags

Content: Topics

● Latent Dirichlet Allocation (LDA)○ R package: topicmodels○ Optimal # of topics: 34

● Make sense of topics○ R package: LDAvis○ Interactive exploration, clustering

● Track selected topics

#LAK15Meta

Select a cluster

Select a topic

Select a term

Types of Topics

1. Information-sharing related to conferences and the community

2. Experience-sharing and comments3. Specific research topics (e.g., MOOC,

assessment, students, course design)

#LAK15Meta

Change with Topics

Summary

● An extended reach and increasing interactions● Denser, more reciprocal networks● Peripheral and in-persistent participation● Emergence of multiple sub-communities● Diverse & fluctuating research topics

#LAK15Meta

Limitations & Future Work

● Representativeness of the LAK community● Potential loss of (earlier) data● Challenges posed by briefness of a tweet

● Combine tweets and academic publications● Connect/compare tweeters with authors/attendees● Compare with other closely related communities (e.

g., EDM, LS)● Dive into chains of conversation

Collaborative #LAK15Meta

Thank You!@bodong_c

[email protected]://meefen.github.io/

Special thanks to Martin Hawksey & all LAK tweeters!

#LAK15Meta

#LAK15Meta